Question: What patients do you help with your patterns?

Sreejita Ghosh answered on 15 Jun 2020:

There are 2 patient groups I try to help with my algorithm. First group I help diagnose faster are babies and neonates who have hormone/steroid imbalance leading to them have life-threatening conditions requiring early intervention with the required hormone therapy. I get hormone profiles of healthy subjects and patients from my doctor friend Liz. My algorithm “learns” the typical ‘relative’ hormone pattern for healthy and each of the hormone disease condition. By ‘relative’ I mean how much hormone A is wrt hormone B or C or how much hormone B is wrt C or E, and so on, because each of us might have different absolute amounts of hormone. What causes the hormone imbalance related diseases are how the relative amounts of hormone differ. Anyway, so after the algorithm (LVQ) learns typical pattern of each condition when hormone profile of a new subject is presented, LVQ finds which typical pattern this new profile is closest to and assigns it the label of that typical profile.

The second group of patients I try to help are those suffering from asthma. In this project I am trying to identify the different stages of asthma patients under treatment. The algorithm use is from the same family- LVQ.

My patterns not just help identify which subject is which patients, but it also helps in learning about each of the disease conditions.